Machine learning aided meta-analysis of the impacts of non-biodegradable and biodegradable microplastics on soil microbial communities

Abstract

Microplastics (MPs) have emerged as significant pollutants in terrestrial ecosystems. However, their impacts on soil microbial communities remain poorly understood. In this study, a comprehensive meta-analysis has been conducted, integrating 1220 paired observations from 64 publications, with a particular focus on comparing the effects of non-biodegradable (Non-bio) and biodegradable (Bio) MPs. Additionally, a machine learning approach has been developed to predict these impacts and identify key contributing factors. Our dual-method approach enables a more precise and comprehensive assessment of MPs’ ecological consequences in soils. The findings revealed that Non-bio MPs reduced microbial diversity by 6.95 % but increased microbial biomass and altered community structure by 15.05 % and 55.76 %, respectively. In contrast, Bio MPs amplified these effects, increasing microbial biomass and community structure by nearly 3.4-fold and 4-fold, respectively. Notably, microbial functions increased by 7.52 % under Bio MPs, whereas Non-bio MPs showed no significant impact. Boosted Regression Tree (BRT) analysis identified soil properties (TN, TC, SOC, pH) and MPs characteristics (polymer type, size and concentration) as key drivers of microbial responses. Although Random Forest models achieved reasonable accuracy in predicting the impacts of MPs on microbial diversity and community structure, they performed poorly in predicting microbial functions due to complex and varying enzyme responses. This study highlights the importance of MP biodegradability and underscores the need for longer-term research and comprehensive risk assessments. Future work should prioritize expanded datasets and advanced modeling techniques to unravel the intricate interactions between MPs and soil microbial communities, ultimately supporting more sustainable environment management strategies.